- Home
- Skills
- Technical Documentation
- Architecture Map with Real Progress (C4)
More screenshots
Works with the AI Agents you already use
Architecture Map with Real Progress (C4)
Static architecture diagrams go stale the week they are drawn, and progress reports are usually someone's guess.
$9.99
Architecture Map with Real Progress (C4)
Example session with this skill installed
Map the architecture of this repository and show me where each part stands. I want one page I can open on my phone and share with the team.
- Read your context and instructions
- Compiled the architecture map
Summary
Quick mode on the public
Full Stack FastAPI Template repository, then full mode on the bundled example project to show progress and scope. Both outputs below are real script output.
1. Quick mode: Full Stack FastAPI Template
Command: python3 scripts/build_map.py quick . --name "Full Stack FastAPI Template"
page: docs/architecture-map.html
parts: 10 · arrows: 2 · code files: 159 · computed: f4bcfab, 2026-11-01
| Part | Level | Tech | Lines | Tests |
|---|---|---|---|---|
| backend | container | Python, Shell | 1,237 | 58 |
| backend / app | component | Python | 1,209 | 0 |
| backend / scripts | component | Shell | 28 | 0 |
| frontend | container | TypeScript | 8,739 | 61 |
| packages | container | TypeScript | 349 | 0 |
| packages / emails | component | TypeScript | 163 | 0 |
| packages / ui | component | TypeScript | 186 | 0 |
| scripts | container | Python, Shell | 206 | 0 |
| hooks | container | Python | 6 | 0 |
| Total | 10,537 | 119 |
Arrows "uses" found from imports: backend tests → backend/app, emails → ui.
What quick mode cannot show: progress, status, people and external systems. For those, run full mode: --write-spec docs/ARCHITECTURE-MAP.md writes a spec draft with the 10 parts and the module map it found, ready for review.
2. Full mode: bundled example (Online bookshop, fake data)
Command: python3 scripts/build_map.py example --out example-map.html
| Part | Status | Progress | Note |
|---|---|---|---|
| Website | done | 100% | |
| Service | in progress | 35% | critical area |
| Service / Catalog | in progress | 60% | |
| Service / Orders | in progress | 40% | 1 task with PR open, 1 defect reproduced |
| Service / Users | blocked | 22% | waits for decision D-1 (two-factor login) |
| Service / Payments | to do | 0% | |
| Mobile app | out of scope | dimmed, still listed |
Total for the site: 12 done · 12 designed · 10 to design, out of 34 work items. Unassigned: OP-9 Electronic invoices (no part named).
Simulate exclusion on Payments: 1 part excluded · 2 parts break (Orders and Mobile app depend on it) · 7 work items leave scope. Nothing is saved; "Copy decision" gives you the text to approve.
Next steps
- Open
docs/architecture-map.htmlin any browser, phone included. - Review the spec draft and approve the parts, statuses and scope the agent proposes.
- Rebuild with
build_map.py full . --spec docs/ARCHITECTURE-MAP.md; use--checkto know when the page is out of date.
Connects securely to your tools. The creator never sees your data.
About this skill
Static architecture diagrams go stale the week they are drawn, and progress reports are usually someone's guess. Architecture Map turns a code repository into one self-contained HTML page that shows what the project is made of, how the parts depend on each other and how far each part really is. Every number on the page is computed by a bundled Python script from the repository and a written spec, with the commit and date it was computed from. Nothing is estimated by the model.
What you get
- Architecture view: a zoomable C4-style map (context, containers, components). Zoom, pan and pinch on phones, a lens to go one level down, a breadcrumb to go back up, arrows that stay attached at every zoom level.
- Progress view: nested tiles colored by share done. Containers sum their parts, never average them.
- Scope view with Simulate exclusion: tick the parts you want to postpone and see which other parts break along the written dependencies and how many work items leave scope. It copies a decision text for you to approve; nothing is saved.
- Roadmap view: steps ready to start and steps waiting, each open step with a "Copy prompt" button.
- Attention view: tasks in progress, tests exposing a defect, open decisions, stalled parts, unassigned items.
- Details panel: status, progress, links and every work item of a part and of the parts inside it.
- Download image: SVG of the current view, ready for a deck or a ticket.
Light and dark follow the system. Status is always shown as color, icon and word, never color alone. The page works on desktop and phone and needs no server.
Two modes
| Quick | Full | |
|---|---|---|
| Needs | the repository | the repository and a spec file |
| Parts | code folders | the spec's Parts table |
| Arrows | imports (Python, JS/TS) | imports plus written links |
| Numbers | lines, tests, TODO/FIXME | plus progress bar, status, decisions, open points, defect tests |
| Time | one command | a short guided setup, one approval per phase |
Start with quick mode to see the map in one command. Move to full mode when you want progress, scope and roadmap: the script writes a spec draft from what it found, and the agent proposes each choice for your approval before writing it.
What is inside
SKILL.md: the rules the agent follows (principles, both modes, calculation rules, update flow, limits).scripts/build_map.py: the generator. Python 3.8+, standard library only. Reads the repository and the spec, counts lines and tests, applies the calculation rules and writes the page.--checktells you whether the page is up to date.assets/map.html: the page. It is the same for every project and is never edited per project.references/format.mdandreferences/example.json: the data format and a fake project to try the page.
Why it is different
- Written rules, generated data. The map is a spec plus data computed from it. Lose the page and you lose nothing: rebuild it.
- No invented parts or arrows. A part with no known work says "not broken down yet", never 0%. Gaps are shown to you, never filled in.
- Same result for every buyer, fewer tokens. The script does the counting, so the agent does not rewrite a generator for each project.
Limits
- Import arrows are detected for Python and JS/TS; other languages get arrows from the spec's Links table.
- Test counts use common test patterns per language; unusual frameworks may count zero.
- Level 4 (operations) is supported by the page and the format but not generated by the script.
Icons: Lucide (ISC). Colors: GitHub Primer (MIT).
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Technical Documentation.
Auto-imports support Python/JS/TS only. Other languages need manual Links table entries. Level 4 generation is not automated. Python 3.8+ required for the build script.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 16 days ago
- Software
Python 3.8+
Permissions
File Scopes
Reads the repository to count lines, tests and imports. Writes only the map page and the spec file (by default docs/architecture-map.html and docs/ARCHITECTURE-MAP.md). Runs the bundled Python script from the shell. No network access.